Find out if ChatGPT recommends you

When someone asks ChatGPT for “a good family dentist near Round Rock” or asks Perplexity which espresso grinder to buy, the answer usually names a few businesses and cites a few websites. AI visibility tracking measures whether your site is one of them. This article explains exactly how RankThat measures it, what the numbers on the screen mean, where the method has limits, and what you can actually change.

In short: RankThat checks Google AI Overviews, ChatGPT, Perplexity, Gemini and Claude. It records when an answer cites your site, where in the list of sources you appear, and which competitors are cited instead. It then combines that with an on-site check of how readable your site is for AI crawlers into one AI Visibility Score, with specific fixes underneath.

Two kinds of scan

The page lives under AI SEO › AI Visibility. Below the score card are two scan buttons, because the assistants are checked in two different ways.

The first scan finds questions where you already appear, which is a good way to discover what assistants associate you with. The second asks specific questions fresh, so you see today’s answer.

Where the questions come from

You don’t type the questions in. For the live scan, RankThat picks them in this order: questions where Google or ChatGPT already cite you, then your project’s tracked keywords, then the titles of topics you’ve approved in the content planner. That keeps the questions tied to what the business sells. The Overview tab also shows Suggested Queries to Track, related questions and keywords you aren’t covering yet, as ideas for content.

Mentions, citations and position

These three words get used loosely, so here is what they mean on this page:

• A mention is an AI answer RankThat has recorded for one of your questions.
• A citation means your domain or one of your pages appears in that answer’s sources. This is the number that matters most: it is the assistant pointing people at your site. The glossary entry on LLM citations explains why.
• Position is where you appear in the list of sources. First of five is better than fifth of five.

The Overview tab shows total mentions, cited mentions with your citation rate, AI search volume, average citation position and a sentiment score, then a trend chart, a breakdown by platform, your top AI keywords and your most cited pages. The other tabs are Competitors, Keywords, Cited Pages and Mentions, where you can read the answers themselves.

How the AI Visibility Score is built

The score runs from 0 to 100 with a letter grade (A from 90, B from 80 and so on). It has five parts:

If a part has no data yet, it is left out and the others are reweighted, so a new project isn’t marked down for scans that haven’t run. The score is saved daily, giving you up to 90 days of trend.

The AI readiness part matters because it is the one you control directly. Assistants can only cite pages they can read.

From score to fixes

Under the score, RankThat lists up to six recommendations. Some have a Fix with AI button that drafts the change and, on WordPress with the Bridge plugin, applies it after you approve:

• Unblock AI crawlers in robots.txt. Many sites block GPTBot or ClaudeBot by accident through a security plugin. Check yours with the free AI Crawler Checker.
• Add Organization schema, so assistants have clean facts about who you are, where you are and what you do.
• Add an llms.txt file, a plain-text map of your important pages. The llms.txt generator shows what one looks like.

Other recommendations point to content. The “Find topics” link opens Topic Ideas in the content writer, because the most reliable way to be cited for a question is to have the clearest page answering it.

A worked example: Brightside Dental

Brightside Dental runs the Google + ChatGPT scan for the first time. It finds 34 recorded answers mentioning the practice, but only 6 that cite its site. Most citations go to a directory listing and to a competitor’s “cost of dental implants” page.

The score is a C. AI readiness is dragging it down: the audit shows their security plugin blocks AI crawlers, and there is no Organization schema. The agency fixes both with Fix with AI in a few minutes.

Next, they look at which questions the competitor gets cited for. Three are about implant costs and recovery time, topics Brightside has no page on. They approve those topics, write the pages and publish them. Approved topic titles are one of the sources for live-scan questions, so the next Perplexity, Gemini and Claude scan can ask about implant costs directly and show whether the new pages are being cited.

None of that is guaranteed, and nobody can promise an assistant will recommend a business. What the tracking does is show where you stand and whether the work moved it.

Reading the Cited Pages tab

The Cited Pages tab is the most practical view on the page. It lists which of your URLs assistants cite and how often. Three patterns come up again and again:

• The home page is cited for everything. Assistants know the business exists but have no specific page to point to for specific questions. That usually means service pages are thin or missing.
• One old blog post is cited far more than anything else. It probably answers a common question plainly, with numbers. That is the format to repeat.
• Pages you care about are never cited. Check they are allowed for AI crawlers, that they load without scripts hiding the text, and that the answer appears in the first few paragraphs rather than below a long introduction.

Compare that with the competitors’ cited pages in the Competitors tab. If a rival’s “how much does X cost” page is cited and you have no page on cost, the gap is content, not technical.

How often it runs

• Each platform can be scanned by hand once every seven days.
• The Google + ChatGPT scan also runs automatically on the 1st of each month for sites that haven’t been scanned in the previous 20 days, so the monthly report has current numbers.
• Professional plans include four AI tracking runs a month; Enterprise is unlimited. See pricing.
• An alert fires when a platform’s citation count changes.

Honest limits

AI tracking is a young field, and the method has edges you should know before you put numbers in front of a client.

• You can’t add your own custom questions yet. Questions come from the sources described above.
• Location is country-level. A scan can’t pretend to be a searcher in one particular suburb.
• Brand-name mentions without a link aren’t detected on the live Perplexity, Gemini and Claude scans. Those scans count citations.
• Competitor data comes from the Google and ChatGPT scan, not the live scans.
• Sentiment is a simple word check on the sentences around your domain, not a judgment of tone. Treat it as a rough signal.
• Assistants change their answers often. One scan is a sample. The trend over months is what to report.

What to tell clients

Clients ask whether ChatGPT recommends them, and the honest answer has two parts. First, what the data shows today: how often you are cited, by which assistants, for which questions. Second, what you control: whether AI crawlers can read the site, whether the facts are clear, whether the pages answer real questions. Report the trend over months, and avoid promising a recommendation. No one can.

Fitting it into a monthly routine

For the background on why this matters, read what AI visibility means and the guide on how to get recommended by ChatGPT.

Tip: open the Mentions tab and read ten actual answers before changing anything. You’ll learn more about how assistants describe your market from the answers than from the score.